The Romani language, and where it came from
A thousand years of travelling, recorded in grammar. Romani is the receipt for a journey nobody wrote down.
For most of European history, nobody knew where Roma had come from, including in many cases Roma. The guesses were confident and wrong: Egypt, which is where "gypsy" comes from, or Bohemia, which is where the French bohémien comes from. It was settled in the 1760s, and it was settled by accident.
A Hungarian theology student named István Vályi, studying in Leiden, fell into conversation with three students from Malabar in south-west India. He noticed that a great many of their words were words he already knew, from the Roma living near his home town. He wrote a list. The list was the answer.
The evidence is in the vocabulary
Romani is an Indo-Aryan language, a cousin of Hindi, Punjabi and Gujarati. The core is unmistakable once you look. Pani is water; in Hindi, pani. Nak is nose; nak. Dand, tooth. Trin, three. Phral, brother, from the same root as Sanskrit bhrata, and as English brother, at a much greater distance.
What was picked up along the way maps the route as precisely as any document. There is a layer of Persian, a layer of Armenian, and a very heavy layer of Greek: drom for road, foros for town, and the way numbers and grammar behave in places. Byzantine Greek that thick means centuries in Byzantine territory, not a season passing through. After that the language splinters, and each dialect takes on whatever it settled beside: Romanian, Hungarian, Serbian, Turkish, Slovak, Spanish, German.
No one recorded the migration. The language recorded it anyway.
One language or many
Both, depending on who is asking and why. A speaker of Kalderash and a speaker of Sinti can find each other slowly, or not at all. There are Vlax dialects, Balkan, Central, Northern, and a great many divisions inside each. There is no standard Romani and no academy, and repeated attempts to build one have run into the reasonable objection that whoever standardises it decides whose grandmother was speaking it wrong.
Some varieties are not fully Romani any more. Angloromani in Britain and Caló in Spain are the vocabulary carried inside English and Spanish grammar, which is what remains after generations of pressure. They are still ours. They are also a warning about what the pressure does.
Where it stands
Something between three and a half and seven million speakers, which is a wide range because nobody has ever properly counted and a good many people have excellent reasons not to answer that question on a form. It is recognised as a minority language in a number of European countries. It is taught in very few schools.
The pattern most families would recognise: fluent grandparents, competent parents, children who understand and answer in the country's language. That is how a language ends, and it does not take long. Three generations is usually enough.
It is also entirely reversible, and cheaply. A language survives on being used for ordinary things by people who are not being examined on it. Not lessons. Just use.
ROMANIYA